/AuAc

A warm rain parameterization using aircraft observation and Machine Learning technique

Primary LanguagePythonOtherNOASSERTION

Chiu_AuAc_Standard_2022

"An updated version of the Chiu_AuAC_2021 parameterization for autoconversion rate and accretion rate using in-situ aircraft observations and Machine Learning techniques"

To find Chiu_AuAC_2021 with Standard() and the Initiation() model described in Chiu et al., 2021, select the branch named Chiu_AuAc_2021 (use the drop down menu located at the upper left corner of the page).

Release note

We welcome any feedback and comments

A Fortran version of the code is planned to be released in January, 2023

  • v1.0.0 (09/09/2022): a Python version using Keras-Tensoflow backend of the package is released
  • v2.0.0 (expected on 01/01/2023): a Fortran version of the pacakge will be published

Citation & Contacts

Citation: Chiu, J. C., C. Kevin Yang, Peter Jan van Leeuwen, Graham Feingold, Robert Wood, Yann Blanchard, Fan Mei, and Jian Wang (2020): Observational constraints on warm cloud microphysical processes using machine learning and optimization techniques. Geophys. Res. Lett. doi:10.1029/2020GL091236

  • PI: Dr. Christine Chiu: Christine.Chiu@colostate.edu
  • Co-I: C. Kevin Yang: yang0920@rams.colostate.edu for any issues related to the source code

Descriptions of Machine-Learning models

PLEASE PAY ATTENTION TO THE UNITS!!!

  • Inputs:

    • qc: cloud water content in g/m3
    • Nc: cloud droplet number concentration in /cm3
    • qr: drizzle watar content in g/m3
    • Nr: drizzle drop number concentration in /cm3

  • Outputs:

    • Pau: autoconversion rate in g/cm3/s
    • Pac: accretion rate in g/cm3/s

About this package

This package comprises 5 parts:

  1. Chiu_AuAc_Standard_2022_model.hdf5:

    • Contain the weights and the biases for the Artificial Neural Network (ANN) for the Chiu_AuAc_Standard_2022() model.

    • Used in Chiu_AuAc_Standard_2022_module.py

    • Available in this Github repository.

  2. Scaler.mat:

    • Contain the scaling used to normalize the INPUT and OUTPUT variables in the training dataset.

    • Used in Chiu_AuAc_Standard_2022_module.py

    • Available in this Github repository.

  3. Chiu_AuAc_Standard_2022_module.py: PLEASE DO NOT MODIFY THE CONTENT

    • This is a python module that contains one function: Chiu_AuAc_Standard_2022(). The function performs the following tasks:

      • Inititialze the Artificial Neural Network (ANN) with the trained weights and bias loaded from the Chiu_AuAc_Standard_2022_model.hdf5 file

      • Scale the "Input_Data" with the scaling information obtained from the training dataset

      • Make predictions with the Artificial Neural Network (ANN)

    • Available in this Github repository.

  4. ExampleDatad.mat:

    • This example dataset will be used in run_example.py for predicting Pau and Pac

    • Available in this Github repository.

  5. run_example.py:

    • An example script that demonsrates how to run the Chiu_AuAc_Standard_2022() model with the ExampleData.mat.

    • Available in this Github repository.

Installation (from scratch)

  • Step 0: for v1.0.0, this package can run on Windows, MacOS, and Linux with a appropriate python package manager installed in the operation system.

  • Step 1: make sure that the following packages are installed in your Python3 environment:

    • Tensorflow

    • Keras

    • Numpy

    • (optional; for run_example.py) SciPy

  • Step 2: put all the necessary files (5 in total) in the working directory; you should have:

    • Chiu_AuAc_Standard_2022_model.hdf5
    • Scaler.mat
    • Chiu_AuAc_Standard_2022_module.py
    • ExampleData.mat
    • run_example.py

    YOU ARE ALL SET FOR RUNNING THIS MODULE

  • Notes:

    • You do not need to have CUDA installed in your operational system to use the pacakge
    • The format of the "Input_Data" should be ndarray (click here to learn more about what is ndarray)